E extracta mathematicae Vol. 16, Núm.

2, 287 – 292 (2001)

Homotopy Theory Induced by Cones

F.J. Dı́az, S. Rodrı́guez-Machı́n∗

Dpto. de Matemática Fundamental, Univ. de La Laguna, 38271-La Laguna, Spain

(Research announcement presented by A. Cegarra)

AMS Subject Class. (2000): 55U35, 18Cxx, 55P05 Received December 23, 1999

0. Introduction

Generally, in homotopy theory a cylinder object is used to define homotopy

between morphisms. A cone object is necessary to build exact sequences of
homotopy groups. So, the factorization through cone objects characterizes
the nullhomotopic morphisms.
Here an abstract homotopy theory based on a cone functor that allows
extend nullhomotopy to homotopy is given. Fundamental properties of the
topological cone to develop homotopy theory are generalized in dual standard
constructions by Huber [4]. Some axioms about cofibrations given by Baues
[1] in categories with a natural cylinder are adapted in the case of a cylinder
whose base is collapsed to a single point. The axiomatic theory obtained in this
way allows one to obtain homotopy groups of based objects and cofibrations.
Also, exact homotopy sequences of these groups can be created.
Principal axiomatic theories based on a cone, as homotopy theories given
by Huber [4], Kleisli [6], Seebach [10] and Rodrı́guez-Machı́n [8], are particular
cases of this theory. Moreover, known homotopy theories are examples of this
axiomatic theory or its dual: the classical homotopy of topological spaces,
pointed topological spaces and chain complexes [5]; projective and injective
homotopy theories of R-modules [3]. Others less known are examples too:
some tensorial homotopy theories and proper homotopy theory of exterior
spaces [2].
∗
Supported in part by the Dirección General de Universidades e Investigación de Canarias
and by the Dirección General de Enseñanza Superior de España.

287
288 f.j. dı́az, s. rodrı́guez-machı́n

1. Category with a natural cone

In this section, one gives the axioms that a category must have to obtain
an homotopy theory based on a cone functor, creating the necessary tools to
develop the homotopy theory and obtaining some basic results.

Definition 1. A C-category, or category with a natural cone, is a cat-

egory C together with a class “cof” of morphisms in C, called cofibrations,
a functor C : C → C, which will be called the cone functor, and natural
transformations k : 1 → C and p : CC → C satisfying the following axioms:
C1. (Cone axiom) p(kC) = p(Ck) = 1C and p(pC) = p(Cp).
f i
C2. (Push out axiom) For any pair of morphisms X ← B ½ A, where i is a
cofibration, there exists the push out square
i -
B > A
f f
? ?
i - X ∪B A
X >
and i is also a cofibration. The cone functor carries this push out di-
agram (called cofibration push out) into a push out diagram, that is
C(X ∪B A) = CX ∪CB CA.
C3. (Cofibration axiom) For each object X the morphisms 1X and kX are
cofibrations. The composition of cofibrations is a cofibration. Moreover,
there is a retraction r for the cone of each cofibration i (r(Ci) = 1). This
latter property is called nullhomotopy extension property (NEP).
C4. (Relative cone axiom) For a cofibration i : B ½ A the morphism i1 =
{Ci, k} : Σi = CB ∪B A ½ CA is a cofibration.
By this definition, isomorphisms and cone of cofibrations are cofibrations.
Moreover the cone functor carries cofibration push outs into cofibration push
outs.

Theorem 1. Given the commutative diagram

f g
Y ¾ X - Z
α γ β
? ? ?
Y’ ¾ X’ - Z’
f0 g0
 homotopy theory induced by cones 289

if α, β, γ and one of the morphisms {g 0 , β} : X 0 ∪X Z → Z 0 or {f 0 , α} :

X 0 ∪X Y → Y 0 are cofibrations, then α ∪ β : Y ∪X Z → Y 0 ∪X 0 Z 0 so is.

Definition 2. A morphism f : X → Y is said to be nullhomotopic if

there exists an extension F of f relative to k, that is, F : CX → Y such that
F k = f . An object X is said to be contractible when 1X is nullhomotopic.
Observe that all morphism factored through a contractible object is null-
homotopic.

Theorem 2. Given a morphism i : B → A, the following sentences are

equivalent:

a) i verifies NEP.
b) Every nullhomotopic morphism f : B → X has a nullhomotopic exten-
sion fe : A → X relative to i.
c) Every nullhomotopic morphism f : B → X has an extension fe : A → X
relative to i.
d) k : B → CB has an extension e
k : A → CB relative to i.

By NEP all morphism with contractible domain or codomain can be ex-

tended relative to any cofibration with the same domain. This will be the
main tool used along this paper to obtain homotopies.

Definition 3. Given a cofibration i : B ½ CA, a morphism f0 : CA → X

is said to be homotopic to f1 : CA → X relative to i if there exists an
extension F of the morphism {f0 pCi, f1 } relative to i1 , that is, a morphism
F : C 2 A → X such that F i1 = {f0 pCi, f1 } : CB ∪B CA → X.
The relative homotopy relation is an equivalence relation compatible with
the composition of morphisms. Given morphisms i : B → A and u : B → X,
Hom(A, X)u(i) = {f : A → X / f i = u} denotes the set of extensions of the
morphism u relative to i.

[CA, X]u(i) = Hom(CA, X)u(i) / '.

If F : f0 ' f1 rel. i then hF : hf0 ' hf1 rel. i.
If Cf i = jg and H : h0 ' h1 rel. j then HC 2 f : h0 Cf ' h1 Cf rel. i.
If Cf i = jg is a push out diagram: h0 ' h1 rel. j if and only if
h0 Cf ' h1 Cf rel. i.
290 f.j. dı́az, s. rodrı́guez-machı́n

2. Homotopy groups

Here, homotopy groups relative to a cofibration or referred to an object

are built, and usual properties of them are given.
Along this section, the C-category will be pointed, that is, ∗X is a cofibra-
tion for all object X and C∗ = ∗, where ∗ is the initial object.
By NEP, given a based cofibration i : B ½ CA, there is µ : C 2 A →
C A ∪CB C 2 A = C(CA ∪B CA) such that µi1 = k(pCi ∪ 1). If F : 0 ' 0
2

rel. i then F = {F, 0}µ : 0 ' 0 rel. i. Moreover, if G : 0 ' 0 rel. i then
F ∗ G = {F , G}µ : 0 ' 0 rel. i.
Using that µ∗ : [C 2 A ∪CB C 2 A, X]0(k) → [C 2 A, X]0(i1 ) is a bijection one
can prove the following

Theorem 3. π1i (X) = [C 2 A, X]0(i1 ) is a group with multiplication and

inverse induced by “∗” and “−”, respectively.

Let us remark that the multiplication on π1i (X) does not depend on the
choice of the extension µ.

Definition 4. The n-th homotopy group of an object X relative to a

i
cofibration i : B ½ CA is defined by πni (X) = π1n−1 (X).

Given a cofibration i : B ½ A, the codomain of the cofibration in is C n A.

In this way, homotopy groups relative to any cofibration i : B ½ A can be
defined for n ≥ 2.
The homotopy group πn∗A (X) is also denoted by πnA (X) and called the n-th
homotopy group of the object X referred to the object A.
is
By Definition 4 πni (X) = πn−s (X). Moreover, all morphism f : X → Y
i i
induces f∗ : πn (X) → πn (Y ), and all commutative square f i = jg induces an
homomorphism (C n f )∗ : πnj (X) → πni (X). If the square is a push out diagram
then (C n f )∗ is an isomorphism.

3. Exact sequences of homotopy groups

Now, the exact homotopy sequence associated to a pair is built.

Let cof C be the full subcategory of Pair C whose objects are cofibrations.
(X, Y ) will denote an object of cof C with associated cofibration f : Y ½ X.

Theorem 4. cof C is a C-category, with C : cof C → cof C defined by

C(X, Y ) = (CX, CY ), where the associated cofibration is Cf , and C(g, h) =
 homotopy theory induced by cones 291

(Cg, Ch); k = (k, k), p = (p, p) and cofibrations defined as those morphisms
(u, v) : (X, Y ) ½ (X 0 , Y 0 ) such that v and {f 0 , u} : Y 0 ∪Y X ½ X 0 are
cofibrations in C.

Definition 5. The (n + 1)-th homotopy group of the pair (X, Y ) relative

i (Ci,i)
to a cofibration i is defined by πn+1 ((X, Y )) = πn ((X, Y )), where (Ci, i) :
(CB, B) ½ (CA, A) with associated cofibrations kB and kA , respectively.

Theorem 5. The following sequence of groups is exact:

f∗ j δ f∗
..... → π3i (Y ) → π3i (X) → π3i ((X, Y )) → π2i (Y ) → π2i (X)

where f∗ is defined in the section 2, j([F ]) = [(F, 0)] and δ([(F, G)]) = [G].

The above sequence is denominated the exact homotopy sequence of (X, Y )

relative to i. If i = ∗A one obtains the exact homotopy sequence of (X, Y )
referred to A.

4. Examples

The developed theory has an obvious dual version with cocones and fibra-
tions. Adjoint functors (cone, cocone) give the same homotopy theory. The
adjunction between the tensorial and Hom functors gives homotopy theories
induced by cones and cocones on the categories of Abelian Groups, R-quasi-
modules (≡ abelian groups verifying the properties of R-modules save the
external associative property) and R-modules:
If X denotes an object in any of such categories, R is an unitary ring
and f : R → S is an homomorphism of unitary rings, then CX = X ⊗Z R,
CX = X ⊗R R and CX = X ⊗R S, respectively; C 0 X = Hom Z (R, X),
C 0 X = HomR (R, X) and C 0 X = HomR (S, X), respectively; k : X → CX is
defined by k(x) = x ⊗ 1 and k 0 : C 0 X → X by k 0 (α) = α(1); p : C 2 X → CX is
defined by p(x ⊗ r ⊗ s) = x ⊗ rs and p0 : C 0 X → C 02 X by p0 (α)(r ⊗ s) = α(rs).
Homotopy theories obtained by Baues [1] on categories with natural cylin-
ders or natural path objects are also induced by cones or cocones, respectively.
In this way, the homotopy theory defined by Kamps [5] on chain complexes of
an abelian category can be induced by cones and cocones; the classical homo-
topy theory of topological spaces and the homotopy theory of exterior spaces
[2] are induced by cones.
292 f.j. dı́az, s. rodrı́guez-machı́n

The classical homotopy theory on pointed topological spaces is also gener-

ated by cones and cocones, where cones are known and cocones are the spaces
of the paths with initial point the base point of the topological space.
The free and Hom (Hom (−, Q1 ), Q1 ) functors on the category of R-mod-
ules are cocone and cone functors that induces, respectively, the projective
and injective homotopy theories defined by Hilton [3], where Q1 is the additive
group of the rational numbers modulo the integers.

References
[1] Baues, H.J., “Algebraic Homotopy”, Cambridge Studies in Advanced Math-
ematics 15 , Cambridge University Press, Cambridge - New York, 1989.
[2] Garcı́a-Calcines, J., Garcı́a-Pinillos, M., Hernández-Paricio,
L.J., A closed simplicial model category for proper homotopy and shape
theories, Bull. Austral. Math. Soc., 57 (1998), 221 – 242.
[3] Hilton, P.J., “Homotopy Theory and Duality”, Gordon and Breach Science
Publishers, New York - London - Paris, 1965.
[4] Huber, P.J., Homotopy theory in general categories, Math. Ann., 144
(1961), 361 – 385.
[5] Kamps, K.H., Note on normal sequences of chain complexes, Colloq. Math.,
39, 2 (1978), 225 – 227.
[6] Kleisli, H., Homotopy theory in Abelian Categories, Canad. J. Math., 14
(1962), 139 – 169.
[7] Kleisli, H., Every Standard construction is induced by a pair of Adjoint
Functors, Proc. Amer. Math. Soc., 16, 3 (1965) 544 – 546.
[8] Padrón, E., Rodrı́guez-Machı́n, S., Model additive categories, Rend.
Circ. Mat. Palermo (2) Suppl., 24 (1990), 465 – 474.
[9] Quillen, D.G., “Homotopical Algebra”, Lecture Notes in Math 43,
Springer-Verlag, Berlin - New York, 1967.
[10] Seebach, J.A. JR., Injectives and homotopy, Illinois J. Math., 16 (1972)
446 – 453.